Build your own agent harness.
Most agent courses teach you the buttons. This one teaches the control system, and you leave holding a harness you wrote yourself: regulator. Each module opens with a regulatory question an autonomous agent forces on you, derives the mechanism that answers it, then shows the Pi feature and the typed treatment underneath regulator, the control plane the course is built against.
Prompts advise. Types describe. Gates enforce.The premise the whole course is organized around
Agents need a regulator, not just a good prompt.
The Viable System Model is Stafford Beer's answer to an old question: what does any system, whether an organism, a company or an autonomous agent, need structurally to stay viable while its environment keeps generating more variety than anyone predicted? Ashby's Law gives the constraint. Only variety can absorb variety, so a regulator with fewer possible responses than the disturbances it faces will eventually fail somewhere it can't see.
A model plus a good prompt is not a regulator. It is a hope, and hope doesn't survive contact with a real codebase, a real filesystem, or real money moving. This course treats the harness itself as the regulator: explicit functions, typed channels, and gates that hold when the prompt is ignored or the model changes, and even when the task has nothing to do with code.
Not a prompt pack, not a persona library.
If you've used an agent for real work and hit its limits, this course gives you the vocabulary and the mechanism to fix that inside a harness you write yourself.
| Who | What this course gives you |
|---|---|
| Senior and staff engineers running agents on real codebases | A principled reason for every guardrail, and a harness tuned to their repository. |
| Platform and DevEx teams | A design method for an internal agent platform, plus the evaluation machinery to prove it helps. |
| Agent-tool builders | Deep Pi extension and SDK competence, and a structural critique of "five agents in a trench coat." |
| Technical leaders evaluating autonomy | A vocabulary for where autonomy is safe, and what evidence would demonstrate it. |
Prerequisites. Comfortable TypeScript, Node 22.19 or later, git, a terminal, a key for at least one model provider, and having used some coding agent for real work. No cybernetics background: the primer teaches every term the lessons rely on, with figures, before any code.
By the end, you can
- Describe an agent harness as a control system: what regulates what, over which channel, with what authority, closing which feedback loop.
- Apply the mechanism hierarchy, type → gate → typed tool → model judgment → prompt, instead of solving every problem with prompt text.
- Extend Pi confidently: typed tools, the tool-call lifecycle, context shaping, per-phase model routing, SDK and RPC, packaging.
- Design work units as contracts, with each decision marked fixed, delegated within bounds, or unresolved.
- Build audit that does not depend on the executor's self-report.
- Keep system identity durable and version-controlled instead of trapped in a context window.
- Decide when a human must be interrupted, and build the channel that does it.
- Measure whether your regulation actually improved outcomes, with a control arm.
The same curriculum, three envelopes.
The repository, the written modules, recorded walkthroughs and graded labs. Ends at reference-build checkpoint 14.
regulator at checkpoint 14Two 90-minute live sessions a week, labs between, design reviews, and a capstone crit.
Modules 01, 02, 03, 09 and 10 only, with the labs run on the team's own repository.
A primer, fifteen modules, six parts.
Every module is a fixed unit: question → concept → Pi mechanism → build step → break it → field study → checkpoint. The "break it" drill is how the regulatory need becomes felt rather than asserted.
The full module descriptions, the premise and the reference build are on the course's public page; the product it builds is regulator.
The pilot is small on purpose.
Tell us which format you want, what you run agents on today, and what went wrong last time. We reply to every request, and we'll say plainly if the course isn't the right fit.